Local-only vs cloud sync

SwarmMemory is local-first. Out of the box, your entire knowledge graph lives on your machine as Markdown files and never leaves it. If you work across more than one device, you can opt into cloud sync to mirror the graph everywhere. Understanding the difference matters mostly for one reason: it determines what consumes Honey.

Local-only (default)

This is the default and it is free.

Private — every memory stays on your machine. Nothing is uploaded.
Free — creating, reading, updating, and deleting memories costs no Honey. Keyword search via search_memories is also free.
Fast — reads and writes hit local files and the local index directly, with no network round trip.

If you never enable sync, you can use the full Markdown-and-wikilinks graph indefinitely without spending a single Honey on memory operations.

Cloud sync (optional)

Cloud sync mirrors your graph across devices so the same memories follow you between machines. It is opt-in and metered.

Multi-device — the same graph on your laptop and desktop, kept in step.
Metered — each sync costs roughly 1 Honey.
Semantic search — meaning-based queries via semantic_search cost roughly 2 Honey per query, because they run an embeddings lookup.

Only two things consume Honey in SwarmMemory: cloud sync (~1 Honey per sync) and semantic search (~2 Honey per query). Everything local — plain reads, writes, and keyword search — is free.

Choosing a mode

If you work on one machine and keyword search is enough, stay local-only and pay nothing for memory. Turn on cloud sync when you genuinely need the graph on multiple devices or want meaning-based retrieval across your notes. You can keep most of your workflow free even with sync enabled, simply by preferring keyword search_memories over semantic_search for routine lookups.

For the full breakdown of every action that draws down your balance, see What consumes Honey.